Alphabet’s Project Loon Helium Balloons in Kenya have gotten a new software to run its navigation system using Artificial Intelligence.
The software will use a set of written and executed algorithms deeply reinforcing a learning based flight control system that is said to be more effective than the older human-made algorithm.
The system is already operational in the Kenya’s Project Loon fleet of Balloons. Loon first launched commercial internet services in Kenya, after running a series of tests. The project aims to connect internet to people living in underserved areas in the country.
Reinforcement learning is an AI technique that allows software to teach itself skills using trial and error. With help from the Google Team in Montreal, Loon’s AI Lab taught the flight control software how to pilot the balloons using computer simulation.
The system was therefore able to improve with time before it was set up on a real functioning balloon.
“While the promise of RL (reinforcement learning) for Loon was always large, when we first began exploring this technology it was not always clear that deep RL was practical or viable for high altitude platforms drifting through the stratosphere autonomously for long durations,” explains Sal Candido, Loon’s chief technology officer said in a blog post.
“It turns out that RL is practical for a fleet of stratospheric balloons. These days, Loon’s navigation system’s most complex task is solved by an algorithm that is learned by a computer experimenting with balloon navigation in simulation.”
Loon says that the Reinforcement Learning type of AI is a first in the commercial aerospace system. It also says that the AI outperforms human intelligence.
“To be frank, we wanted to confirm that by using RL a machine could build a navigation system equal to what we ourselves had built,” Candido writes.
“The learned deep neural network that specifies the flight controls is wrapped with an appropriate safety assurance layer to ensure the agent is always driving safely. Across our simulation benchmark we were able to not only replicate but dramatically improve our navigation system by utilizing RL.”
The company first tested RL in Peru in 2019, pitting it against a traditionally designed human built algorithm called StationSeeker. The AI beat and outperformed the human one paving way for more experiments.
Loon now believes its system can “serve as a proof point that RL can be useful to control complicated, real world systems for fundamentally continual and dynamic activity.”